Statistical Analysis, Feature Selection, and Optimization Techniques for Accurate Seismic Event Forecasting and Risk Assessment in Advanced Earthquake Prediction

Earthquake prediction involves forecasting the location and timing of future quakes. Short-term prediction prompts immediate responses such as warnings and evacuations, while long-term prediction focuses on seismic trends for planning. Despite advances, short-term forecasts remain unreliable but assist in preparedness for high-risk areas. This study applies ML models like QDA, optimized with DOA and HGSO, to enhance accuracy and identify key features. A feature selection method is used to determine the most influential attributes, such as sig, MMI, CDI, and latitude. QDA achieved 0.981 training, 0.933 testing, and 0.971 overall accuracy. The “sig” feature had the highest F-statistic, boosting performance.